This paper reports two experiments concerned with verbal representation in the test stage of recognition memory for naturalistic sounds. In Study I 10 subjects produced verbal description to 50 sounds and after a 7-day delay 5 subjects reviewed self-produced verbal descriptions while 5 did not prior to testing for recognition. Review of descriptions increased recognition to 94.4% compared to 88.5% for those with no review. In Study II 16 subjects produced verbal descripitions to 50 naturalistic sounds. After a 7-day delay all subjects reviewed descriptionss of one-half of the items which were "targets" and 8 subjects reviewed description of 25 "non-targeted" items. The latter condition produced interference, observed as a significant decrement in recognition from 90.3% to 86.5%.